accheng

5 ms·

1.
by accheng 10mo ago
Thanks for the link! I am not familiar with the company but reminds me of the whole formal methods debate in distributed systems. Sure, writing TLA+ specs is the 'correct' deterministic way to build a Raft implementation, but in r
2.
by accheng 10mo ago
Interesting, did you have any learnings that would apply to this problem now?
3.
by accheng 10mo ago
Check out our blog series here: ucbskyadrs.github.io!
4.
by accheng 10mo ago
As described in our VLDB paper ( https://www.vldb.org/pvldb/vol17/p2694-cheng.pdf ), we had to implement a number of optimizations and integrate scheduling with CC to get good performance.
5.
by accheng 10mo ago
What did you use to try to optimize scheduling?
6.
Turns Out AI Is Not Good at Database Transaction Scheduling(adrs-ucb.notion.site)
13 points by accheng 10mo ago|10 comments
7.
by accheng 11mo ago
Awesome results! Did you use the simulator from the NSDI paper directly?
8.
by accheng 11mo ago
Neat, thanks for the link! Seems like the agent feedback loop made a big difference for your use case. Would be interesting to see how much of a difference more advanced methods (like the genetic algs you mentioned) would make
9.
by accheng 11mo ago
The algorithm works for MoE load balancing in general
10.
by accheng 11mo ago
We've found that these frameworks do well for systems performance problems and expect that the range of problems for which they apply will increase as they models and frameworks improve. See our paper ( https://arxiv.org&#x2F
11.
by accheng 11mo ago
That's a good point! The load balancing of the original algorithm was already quite good so our goal was to try to get something that could achieve similar results but could run faster since runtime was also a concern.
12.
by accheng 11mo ago
ADRS = AI Driven Research for Systems. See our previous blog post ( https://www.sigops.org/2025/barbarians-at-the-gate-how-ai-is... ) and our paper ( https://arxiv.org/pdf/2510.06189 ) for more detail
13.
by accheng 11mo ago
Thanks for sharing your blog! Very interesting work, 100% agree with your 3 criteria on the sweet spot for AI. Most systems performance problems fit right in
14.
by accheng 11mo ago
The code was quite short and easy to read. Specifying the right scoring function and scoping the problem are key parts of getting good results with ADRS.
15.
by accheng 11mo ago
Yes, getting the right workloads and ensuring correctness are crucial parts of the process
16.
by accheng 11mo ago
This is a great question! By analyzing the logs of OpenEvolve with the full model outputs, we observed how the AI got its ideas (seemed to be pulling from literature in the space) and how it tried to apply them. So in some sense, it "r
17.
by accheng 11mo ago
As an author of the blog, I'll note that this was one of the easiest applications of ADRS. Bowen, who was leading this effort, got things running within a day or two and the initial runs were with free Google credits! It was exciting t
18.
We Asked AI to Design Systems Algorithms. It Beat Us in 12 Hours for <$20(sigops.org)
9 points by accheng 11mo ago|1 comments